Figure 1
SVM regression values for each sample. For each sample, we applied the surgery and stimulation regressions that were trained using large-window pixel-wise data (fiber density, axon packing, and myelin packing). The average output was calculated for each sample with control samples in blue circles, sham samples in green squares, and stimulated samples in orange (with stimulation levels noted in mA; McCreery level; and Shannon k). Fast Fourier transform kernel density estimation was then used for each sample’s resulting 2D multivariate distribution to estimate the distribution’s 1 sigma (68% confidence interval) represented by the solid lines (stimulation samples lines also have associated marker) and each sample distribution’s 2 sigma (95% confidence interval) represented by the shaded regions. Some portions of the control and stimulated sample data (but none of the sham) were inferred to have changes related to high levels of stimulation but not related to the electrode being implanted (concentration of values in the bottom right of the plot). However, no stimulated sample data falls in the low surgery and low stimulation region (bottom left).
